342 Proc. Japan Acad., 57, Ser. A (1981) [Vol.

57 (A),

The r Function of the Kadomtsev.Petviashvili Equation

Transformation Groups for Soliton Equations.
By Masaki KASHIWARA and Tetsuji MIWA
Research Institute for Mathematical Sciences, Kyoto University
(Communicated by K.6saku YOSlDA, M. J. A., Sept. 12, 1981)

The notion of r function was first introduced by Sato, Miwa

and Jimbo in a series of papers on holonomic quantum fields [1].
There the r functions were simply expressed as the expectation values
of field operators which belong to the Clifford group of free fermions.
Then by several authors [2]-[4] the concept was generalized and ex-
ploited in the study of monodromy and spectrum preserving defor-
mations. They also showed that r functions are nothing other than
the dependent variables used by Hirota in this theory of bilinear
equations [5].
This is the first in a series of papers [6], [7] by the present au-
thors, E. Date and M. Jimbo, which aims at a further study of r
functions and soliton equations.
The main results in the present paper are the following, a) We
construct a Clifford operator (x) so that for any even Clifford group
element g the expectation value r(x)= (e(x)g} gives us a solution to the
hierarchy of the KP (Kadomtsev-Petviashvili) equations in Hirota’s
bilinear form. b) Define polynomials p(x) (]=0, 1, 2,...) by
exp Mx j=O
p(x)M.
The KP hierarchy contains the 2ollowing infinite number of bilinear
equations

det Pf2 "Pf2+m-2 z(x). r(x)= O,

(fl_f._" ._fl, m2, D--(DI, D./2, D3/3,...)).

Our work is deeply related to the recent progress [8] by lV[. Sato
and Y. Sato on the structure of r functions for the KP hierarchy. In
fact, our starting point was Sato’s lecture [9] in which he claimed
 _
No.. 7] Transformation Groups for Soliton Equations. I 343

that the space o solutions to the KP hierarchy is neatly parametrized

by the Grassmann manifold of infinite dimensions.
In 1 we give the definition o the r 2unction or the KP hier-
archy. This part is due to an unpublished work of M. Jimbo, M. Sato
and one o the present authors (T. M.). In 2 we give an expression
or the r unctions of the KP hierarchy by means of the Clifford oper-
ators used in [10]. In 3 we give the explicit orms o Hirota’s bilinear
equations or them using a transformation introduced in [8]. In 4,
as a corollary o the operator expression given in 2, we show that
the characters o the general linear group are r unctions 2or the KP
hierarchy; that is one o the astonishing results o M. and Y. Sato
[9].
We thank M. Sato or his explanation on his recent work with
Y. Sato before publication. We benefited rom discussions with M.
Jimbo.
1. We denote by x=(x, x,...) an infinite number o inde-
pendent variables. The hierarchy of higher-order KP equations is a
system of non-linear partial differential equations with unknown unc-
tions b(x)(/=2,3,... ;0<_m<_/-2)obtained as the compatibility
condition or the ollowing system of linear partial differential equa-
tions for a wave function w(x).
(1)
We set
aw(x)
x (a -o-
+ F, w(x) 1=2,3,

(2) (x, k)= ] k’x.

=1
Assume the system (1) is compatible. Then, by a suitable change of
dependent and independent variables, we cn find formal solution
w(x, k) to (1), containing a spectral parameter k, of the form
(3) w(x, k)= (x, k) exp (x, k),
(4)
l=l
A ormal solution is uniquely determined by (1) up to a actor depend-
ing on k but independent of x. I we fix a ormal solution, the r unc-
tion is consistently defined up to a constant actor by the 2ollowing"
(5) (log r)=-- lt(x)

’
3t_(x).
We set =(a/aXl, (1/2)(a/x), (1/3)(a/ax), ...) and define a modified r
unction r[(x) by
( x--2-’
1
"’) e’
Then the system (5)is rewritten as a ormal identity which directly
connects w(x, k) with r(x)
344 M. KASHIWARA and T. Mwn [Vol. 57(A),

(7) w(x, k)=r(x)/r(x).

Note that a change of r(x) by the exponential o a linear unction gives
a multiplication to w(x, k) by a unction in k.
If l= 1 ia (5), we have
(8) b.o(X)---2(/x) log r(x).
Thus the introduction of r(x) by (5) gives a generalization of Hirota’s
dependent variable transformation [5] for the total hierarchy. Namely,
i the system (1) is compatible, r(x)satisfies infinitely many Hirota
bilinear differential equations. Let us count the degree of Dt as 1.
Note that i P(-D)=--P(D) we have P(D)f(x). f(x)=0 or any unc-
tion f(x). The equation of the lowest degree among non trivial ones
is the ollowing [11].
(9) (D+3D--4D,D)r(x).r(x)=O.
In [8], M. Sato and Y. Sato listed the bilinear equations satisfied by
r(x) up to degree 11, and conjectured that the number o bilinear equa-
tions o degree n is equal to p(n-1), where p(n) is the partition unc-
tion; p(n)={(n, ...,n)]n integer s.t. lnn n and

In 3 we propose a refined version o their conjecture by giving

explicit forms of bilinear equations or every degree.
2. Now we prepare some notations on characters o the general
linear group. Following [9] we set x=trace g/1, where g is an ele-
ment o GL(N, C) or a sufficiently large N. For any Young diagram
Y=(f, ...,
f) we denote by Zr(x) its character evaluated at g. This
is a polynomial in x which is independent o the choice o N larger
than f+... +f. We denote by p(x)the /-th symmetric tensor
(i.e. Y=(/))and by q(x) the /-th skew-symmetric tensor (i.e. Y=(1)
=(1...1)). For l0 we understand p(x)and q(x) to be zero then we
have
(10) e (’) kp(x),
and q(- x) (-)pt(x).
Now we denote by and (me Z) ree ermions such that
[, ]=, [, ]+=0 and [, ]+=0. We also use their
Fourier transforms (d= dk/2=ik)"

(11)
()=
+z
, *()= E k -,
+z
We define the Hamiltonian or the x-flows by
(12) H(x)=
ll
and set
No. 7] Transformation Groups ’or Soliton Equations. I 345

We introduce a Clifford operator (x)= e "(x)" by

p(x)= E (- x)(- x),
(k)+*(k).
1- k/k e ‘(x,’)
This operator is essentially the same as the operator used in [11].
Theorem 1. If g belongs to the Clifford group,
(15) :(x)=((x)g}
satisfies the bilinear eqtions of the KP hierarchy. The 1-th modified
r function
(16) r,...(x) e (x,)e- (, )... e (,,)e- (x,; 1)r(x)
satisfies the following MtKP equation (the 1-th modified KP equation)"
(17) q + (D)r,...,(x). r(x) 0.
We shall give below a sketch of the proof o Theorem 1. First
we note that
(18) r,,...,(x) ([...[[ "e"()+(-x)
-
+51(--x)",
+(k)ki ](_), +(k_) k_](_)_, ..,
([, ](_)= [, ]+(l" even), [, (l" odd)).
Then the proo of (17) is done in the ollowing manner. For sim-
plicity, let us assume that l= 1. Noting that
Z(D)
fg f)g-- (Z()f)(z()g) + f(2()g),
(Z()
we have
z(D)( e()( x)" (k)g}. (’e ()" g
( e,()_l(--x)(--x)(--x) (k)g}( .e,(X) g}
(" e"(x)+( x)" +(k)g}(" e"()+_( x)+7(-- x)" g}
+ (" e"(x)+( x)" +(k)g}(" e"()+_(x)+( x)" g}
0.
In the last step we used Wick’s theorem which is valid only i g belongs
to the Clifford group. Noting (7) we can rewrite the bilinear equation
(17) with l= 1 as the linear equation (1) with l= 2. The latter with l 3

.
is obtained from (22) below with l’= 1.
For any Young diagram we set
p,+ (- D/2)p,+ (0/2)... p 1(O/2)
(19) h)(D)=det p+_(-D/2)p(D/2)...p+_(D/2)
Then we have
Theorem 2. The hierarchy of the MKP (l O) equations contains
the bilinear equation
346 M. KASHIWARA and T. MIWA [Vol. 57(A),

(20) h ) (D)r,,... ,(x). r(x)

0
for each Y.
We conjecture that the h/’(D)’s are linearly independent and ex-
’haust all the bilinear equations o the MKP hierarchy. Since the
degree o h /)(D) is f / /f / / 1, the number o degree n bilinear
equations o the MtKP hierarchy is expected to be p(n-l-1). The
case/=0, i.e.
(21) {degree n Hirota’s bilinear equation o the KP hierarchy}
--p(n-- 1),
was conjectured by M. and Y. Sato in [8]. The proo o (21) will be
given in one o2 our 2orthcoming papers [7], in which we discuss the
relation between soliton equations and the Euclidean Lie algebras.
Now we shall give a sketch o the proo o Theorem 2. First we
note the ollowing [8].
Lemma. I f(x) and g(x) satisfy a bilinear equation Q(D)f(x)

Q(D / k, D/ k ,
g(x)=0, then f(x)and g(x) satisfy
) exp (--(D/k / D/(2k)/
Applying this lemma repeatedly to (17) we find that
)/2)f(x) g(x)-O.

Q(D kt, , .,
k)v,...,,(x), v(x)----- 0
where
Q(D k, , ., k) 1
k,
q (/ + k, +. + k,
k
D + k"2+ + k
2’ ...)
1 1
k,---- /.../-k, ) /+ (-k,+
1
exp (_{( 1 +.../
__) /)/....})
Setting y-(k+... + k)/2] nd taking the limit l--c we obtain
(22)
j=(}
p(--2y)p+, /(Dz) exp yDx ’[I],...,[I,J(X)’’(X)-’--O.
Using Weyl’s character ormula we find the coefficient o Zr(Y) in (22)
to be h’/)(D).
4. In order to compute (?(x)g} explicitly we fix the expectation
value. The ollowing choice o the vacuum was suggested by M. Sato.
(23) lvac=O n_-l, vacl=O n_O,
*nlVac=O n_O, vacl*=O n_-l.
We set
(24) p;(x)= pt_ (-- x)p,_(x)= ((x),*(x)}
1E -1
(Z, , e Z).
Then we have
(25) p(x / y)-- (- x)* (- x)p;(y).
If/_-1 and ,_0, (-)"-p;(x) is the character or Y=(,/I, 1 -") and
No,. 7] Transfo,rmatio,n Gro,ups for Soliton Equations. I 347

(*. (- x)+} (+(x)+*. } -p_./(x). Now let Y be a Young diagram

o the ollowing orm:
+1

The character Z(x) is written as

()/’"/z(x) ((x)+.*.. .....}=(-)
* det (pL(x)),;,...,.

Errata (Proc. Japan Acad., 56A, 9 (1980)). M. Jimbo and T. Miwa: Studies
on Holonomic Quantum Fields. XVII.
p. 405, 1. 9 from bottom: For "(sinh
read "(sinh 2/LE1 sinh 2/LE.)’=(sinh 2/+E1 sinh 2/E)-. 1. 2 from bottom: For
1)---- a---((t + 1)--- a)
d
"N.((t--1) a) ---d{((t
da da \,,

read
da da
)h---o -1/4)( t-d- a)."
da da
,,N((t_ 1)--- a)
p. 406, 1.20" "fl-)=ao/t--l(F(--1/2, 1/2, 1; 1/t)--F(1/2, 1/2,
For 1; l/t).)" read
"fl0(-)=--a0/)--/---l(F(--1/2, 1/2,1; 1/t)--F(1/2, 1/2, 1; l/t))." 1. 11 from bottom"
For "(sinh/LE sinh/LE.) -1’’ read "(sinh 2/LE sinh 2/LE.)-I.
p. 407, 1. 2: For "sinh fl+E1 sinh fl+E" read "sinh 2fl+E sinh 2/+E.."
The authors are grateful to Profs. B.M. McCoy and J. H. H. Perk for point-
ing out these errors.

References
1] Sato, M., Miwa, T., and Jimbo, M.: Publ. RIMS, Kyoto Univ., .14, 223
(1978); IS, 201, 577, 871 (1979); 16, 531 (1980).
2 Okamoto, K.: Proc. Japan Acad., 36A, 264, 367 (1980).
3 Jimbo, M., Miwa, T., and Ueno, K.: Monodro.my preserving defo.rma.tion
of linear ordinary differential equations with rational coefficients I (to
appear in Physica D).
4 Jimbo, M., and Miwa, T.: ditto. II, III (to appear in Physica D).
5 Hirota, R.: Lect. Notes in Math., vol. 515, Springer (1976).
6 Date, E., Kashiwara, M., and Miwa, T.: Transformation groups for soliton
equations II. RIMS preprint, no. 357 (1981).
7 Date, E., Jimbo, M., Kashiwara, M., and Miwa, T. Transformation groups
for soliton equations III-VII. RIMS preprint, nos. 358-362 (1981).
[8] Sato., M., and Sato (Mori), Y.: Kokyuroku RIMS, no. 388, 183 (1980.);
no. 414, 181 (1981) (in Japanese.).
9 Sato, M.: A lecture delivered at the symposium on Random Systems and
Dynamical Systems held at RIMS, Kyoto Univ. on Jan. 7-9, 1981.
[10] Miwa, T.: Clifford ope.rators and Riemann’s monodromy problem (to ap-
pear in Publ. RIMS).
[11] Satsuma, J.: J. Phys. Soc. Japan, 40, 286 (1976).